google / google/ml-compiler-opt
Use same actor kind when BC and training
Open
- Dominant language
- Python
- Stars
- 798
- Forks
- 123
- Avg merge
- 55m
- Merged PRs (30d)
- 4
Description
The BC policy is trained using a QNetwork actor, which is different from the one we use for normal training. See for example:
https://github.com/google/ml-compiler-opt/blob/9d00bcf673109e46b3f5aa6f7a586c4a124c29b6/compiler_opt/rl/inlining/gin_configs/behavioral_cloning_nn_agent.gin#L19
https://github.com/google/ml-compiler-opt/blob/9d00bcf673109e46b3f5aa6f7a586c4a124c29b6/compiler_opt/rl/inlining/gin_configs/ppo_nn_agent.gin#L35
Contributor guide
Assessment
This issue has not been assessed yet.